28 October 2017
R to visualize your data
Why visualize your data?
- summarize information
- check assumptions validity
- many pieces of information at one glance
- ease the understanding
Why use R to do it?
Workflow
Why use R to do it?
Workflow
Why use R to do it?
Here we'll argue why use R for doing your plots:
- visualizing data
- what makes a good graph (some details)
- work-flow (data are already in R)
- base graph / ggplot2 / web graphs (interacting)
- code / reproducible
Why use R to do it?
Different valuable options
Why use R to do it?
Many, many, many packages available
Why use R to do it?
Why use R to do it?
Few lines of codes
library(raster)
library(mapview)
elvBtn = getData('alt', country='BTN', path='./assets/')
mapview(elvBtn)
Why use R to do it?
Few lines of codes
Why use R to do it?
A large community
Why use R to do it?
Common plots used in meta-analyses
What do we want/need to visualize?
Effect Size (ES)
- log risk ratios,
- log odds ratios,
- …,
- non-standard ES
Studies
- by author
- by region
- by species
Effect size – Funnel plot
http://www.metafor-project.org/doku.php/plots:funnel_plot_with_trim_and_fill
Effect size – Forest plot
Vellend et. al 2013
Effect size – Forest plot
http://www.metafor-project.org/doku.php/plots:forest_plot_with_subgroups
Effect size – Radial plot
Effect size – L'Abbé plot
Number of Study & effect size
Vellend et al. 2013
Number of Study & geography
Vellend et al. 2013
Common plots used in meta-analyses with R
R packages for meta-analyses
install.packages("ctv"); ctv::install.views("MetaAnalysis")
Meta vs Metafor – briefly
Meta vs Metafor – briefly
Meta vs Metafor – support
Guido Schwarzer author of
metaSchwarzer, Guido, James R. Carpenter, and Gerta Rücker. Meta-Analysis with R.
Wolfgang Viechtbauer author
metafor
Meta – workflow
- meta:
metabin()
Arguments:
event.e: Number of events in experimental group.
n.e: Number of observations in experimental group.
event.c: Number of events in control group.
n.c: Number of observations in control group.
Meta – workflow
meta:
- `metabin()` / `metagen()`
Arguments:
TE: Estimate of treatment effect.
seTE: Standard error of treatment estimate.
Meta – workflow
meta:
- `metabin()` / `metagen()` - `funnel`, `forest`, ..
install.packages('meta')
library(meta)
data(Olkin95)
# ?Olkin95
meta1 <- metabin(event.e, n.e, event.c, n.c,
data=Olkin95, subset=c(41,47,51,59),
sm="RR", method="I",
studlab=paste(author, year))
forest(meta1)
Meta – workflow
| author | year | event.e | n.e | event.c | n.c |
|---|---|---|---|---|---|
| Fletcher | 1959 | 1 | 12 | 4 | 11 |
| Dewar | 1963 | 4 | 21 | 7 | 21 |
| Lippschutz | 1965 | 6 | 43 | 7 | 41 |
| European 1 | 1969 | 20 | 83 | 15 | 84 |
| European 2 | 1971 | 69 | 373 | 94 | 357 |
| Heikinheimo | 1971 | 22 | 219 | 17 | 207 |
Meta – workflow
Metafor – workflow
Metafor – workflow
Calculate effect size:
escalc()Perform analysis:
rma()Plots:
funnel(), 'forest()', 'labbe()', …
Metafor – workflow
install.packages('metafor')
library('metafor')
Metafor – example
r<-c(0.5,0.6,0.4,0.2,0.7,0.45); n<-c(40,90,25,400,60,50); studynames<-c(1,2,3,4,5,6); b<-data.frame(r,n,studynames) eszcor <- escalc(measure="ZCOR", ri=r, ni=n, data=b) FEmodel<-rma(yi=yi, vi=vi, data=eszcor, method="FE") #forest plot fixed-effect model: pooled effect size, forest(FEmodel); forest(FEmodel,transf=transf.ztor)
Funnel plot – version 2
Funnel plot – version 3
Forest plot
Pros and Cons
Pros
- do everything for you
- good-looking plots with a few line of plot
Cons
- standard ES
- not fully customizable
Custom your plot
graphics or grid?
Few tips
par()plot(),barplot(),funnel()
npt <- 15 vecX <- 1:npt vecY <- runif(length(vecX))
plot(vecX, vecY)
plot(vecX, vecY, xlim=c(0,15)+.5)
plot(vecX, vecY, xlim=c(0,15)+.5, pch=19)
plot(vecX, vecY, xlim=c(0,15)+.5, pch=19, col="#0f7aa2")
plot(vecX, vecY, xlim=c(0,15)+.5, pch='S', col="#0f7aa2")
plot(vecX, vecY, xlim=c(0,15)+.5, type='h', col="#0f7aa2")
plot(vecX, vecY, xlim=c(0,15)+.5, type='h', col="#0f7aa2", lwd=12)
par(lend=2); plot(vecX, vecY, xlim=c(0,15)+.5, type='h', col="#0f7aa2", lwd=12)
par(las=1) plot(vecX, vecY, xlim=c(0,15)+.5, pch=19, col="#0f7aa2")
par(las=1) plot(vecX, vecY, xlim=c(0,15)+.5, pch=19, col="#0f7aa2") abline(v=c(0.5,15.5), lty=2)
par(las=1, xaxs='i') plot(vecX, vecY, xlim=c(0,15)+.5, pch=19, col="#0f7aa2") abline(v=c(0.5,15.5), lty=2)
vCol <- rep(1, npt) vCol[c(2,5,8,9)] <- 2 vSize <- 1+2*runif(npt)
par(las=1, xaxs='i')
plot(vecX, vecY, xlim=c(0,15)+.5, pch=19, col=c("#0f7aa2", '#a70d72')[vCol])
par(las=1, xaxs='i')
plot(vecX, vecY, xlim=c(0,15)+.5, pch=19, col=c("#0f7aa2", '#a70d72')[vCol], cex=vSize)
par(las=1, xaxs='i', mfrow=c(2,2))
for (i in 1:4){
plot(vecX, vecY, xlim=c(0,15)+.5, pch=19, col=c("#0f7aa2", '#a70d72')[vCol], cex=vSize)
}
par(las=1, xaxs='i', mfrow=c(2,2), mar=c(2,2,1,1), oma=c(2,2,0,0))
##--
plot(vecX, vecY, xlim=c(0,15)+.5, pch=19, col=c("#0f7aa2", '#a70d72')[vCol], cex=vSize,
ann=F)
##--
plot(vecX, vecY, xlim=c(0,15)+.5, pch=19, col=c("#0f7aa2", '#a70d72')[vCol], cex=vSize,
ann=F)
##--
plot(vecX, vecY, xlim=c(0,15)+.5, pch=19, col=c("#0f7aa2", '#a70d72')[vCol], cex=vSize,
ann=F)
##--
plot(vecX, vecY, xlim=c(0,15)+.5, pch=19, col=c("#0f7aa2", '#a70d72')[vCol], cex=vSize,
ann=F)
##--
mtext(1, text='Time', outer=T)
par(las=0)
mtext(2, text='Variable', outer=T)
Example funnel plot
Questions?
Let's practice
Reproduce 2 figures from Vellend et al. 2013
Data available on line as well Dataset S1
Figure S2
Vellend et al. 2013
Figure 3
Extra
WordCloud
library(wordcloud2) wordcloud2(demoFreq) wordcloud2(demoFreq, size = 2)
https://github.com/inSileco/wordcloud/blob/master/ourWordcloud.R
WordCloud
Network of co-authors
igraph
Resources
Useful links
2 books:
Chen, Ding-Geng, and Karl E. Peace. Applied Meta-Analysis with R. Chapman & Hall/CRC Biostatistics Series. Boca Raton: CRC Press/Taylor & Francis Group, 2013. + companion website
Schwarzer, Guido, James R. Carpenter, and Gerta Rücker. Meta-Analysis with R. Use R! Cham: Springer International Publishing, 2015. https://doi.org/10.1007/978-3-319-21416-0.